|
On 4/4/06, Todd Lyons <tlyons@xxxxxxxxxx> wrote:
> On Fri, Mar 31, 2006 at 12:35:31PM -0800, Bill Omer wrote:
>
> >I have heartbeat running on two servers. When one dies, heartbeat
> >will run the /etc/init.d/lvs script I wrote to start up the ipvs
> >rules. At the same time, I'm trying to use ipvsadm backup daemon to
> >sync the connections. The problem I'm having is, the connections dont
> >seem to be syncing and when there's a fail over all the established
> >connections are terminated.
>
> Did you ever make any progress on this? I'm not authoritative on it
> because I don't use it, but was thinking of implementing it so would
> like to know what you're finding.
> --
> Regards... Todd
Hi Todd
I think I found what was causing the problem, at least in my situation.
I go rid of the --mcast-interface option, and now all is good. The
downside to this is now I'm not able to do the broadcasts across a
different interface. This may not be a problem for some but may turn
out to be a real issue in my environment. I need to monitor the
broadcast storms ipvs will be issuing across the network.
Below are the commands I'm using. I tested it this morning. I
started 3 session when the 'primary' lvs server had the vip. Then I
shutdown heartbeat so the backup server would take over the vip.
Those 3 connection were not terminated, everything continued to run.
Then I established a fourth connection once the backup lvs server was
running the vip. Now all four connections where running just fine.
Then I started heartbeat again on the primary, the vip was brought
back to the primary and all four connections were still established.
Now I have redundancy between my two directors. All is good. Tonight
I'm going to be testing my lvs setup in a very large environment.
I'll let the list know how that goes.
Regards,
Bill Omer
ipvsadm --start-daemon=master
ipvsadm --start-daemon=backup
|
